BMJ 2014;349:g6015 doi: 10.1136/bmj.

g6015 (Published 27 October 2014) Page 1 of 15

Research

BMJ: first published as 10.1136/bmj.g6015 on 28 October 2014. Downloaded from http://www.bmj.com/ on 1 March 2019 by guest. Protected by copyright.
RESEARCH

Milk intake and risk of mortality and fractures in women

and men: cohort studies
OPEN ACCESS

1 2 3
Karl Michaëlsson professor , Alicja Wolk professor , Sophie Langenskiöld senior lecturer , Samar
3 14 5
Basu professor , Eva Warensjö Lemming researcher , Håkan Melhus professor , Liisa Byberg
1
associate professor
1
Department of Surgical Sciences, Uppsala University, SE-751 85 Uppsala, Sweden; 2Institute of Environmental Medicine, Karolinska Institutet,
Stockholm, Sweden; 3Department of Public Health and Caring Sciences, Uppsala University, Uppsala, Sweden; 4Swedish National Food Agency,
Uppsala, Sweden; 5Department of Medical Sciences, Uppsala University, Uppsala, Sweden

Abstract fracture incidence in women. Given the observational study designs with
Objective To examine whether high milk consumption is associated the inherent possibility of residual confounding and reverse causation
with mortality and fractures in women and men. phenomena, a cautious interpretation of the results is recommended.

Design Cohort studies.

Setting Three counties in central Sweden.
Participants Two large Swedish cohorts, one with 61 433 women (39-74
years at baseline 1987-90) and one with 45 339 men (45-79 years at
baseline 1997), were administered food frequency questionnaires. The
women responded to a second food frequency questionnaire in 1997.
Main outcome measure Multivariable survival models were applied to
determine the association between milk consumption and time to
mortality or fracture.
Results During a mean follow-up of 20.1 years, 15 541 women died
and 17 252 had a fracture, of whom 4259 had a hip fracture. In the male
cohort with a mean follow-up of 11.2 years, 10 112 men died and 5066
had a fracture, with 1166 hip fracture cases. In women the adjusted
mortality hazard ratio for three or more glasses of milk a day compared
with less than one glass a day was 1.93 (95% confidence interval 1.80
to 2.06). For every glass of milk, the adjusted hazard ratio of all cause
mortality was 1.15 (1.13 to 1.17) in women and 1.03 (1.01 to 1.04) in
men. For every glass of milk in women no reduction was observed in
fracture risk with higher milk consumption for any fracture (1.02, 1.00 to
1.04) or for hip fracture (1.09, 1.05 to 1.13). The corresponding adjusted
hazard ratios in men were 1.01 (0.99 to 1.03) and 1.03 (0.99 to 1.07).
In subsamples of two additional cohorts, one in males and one in
females, a positive association was seen between milk intake and both
urine 8-iso-PGF2α (a biomarker of oxidative stress) and serum interleukin
6 (a main inflammatory biomarker).
Conclusions High milk intake was associated with higher mortality in
one cohort of women and in another cohort of men, and with higher
Introduction
A diet rich in milk products is promoted to reduce the likelihood
of osteoporotic fractures. Milk contains 18 of 22 essential
nutrients, including calcium, phosphorus, and vitamin D of
especial importance for the skeleton. Intestinal uptake of these
nutrients is enhanced by the enzymatic capacity to digest lactose
into D-glucose and D-galactose by mutation in the lactase gene,
a variant common in those with northern European ancestry.1 2
An intake of dairy foods corresponding to three or four glasses
of milk a day has been suggested to save at least 20% of
healthcare costs related to osteoporosis.3
A high intake of milk might, however, have undesirable effects,
because milk is the main dietary source of D-galactose.
Experimental evidence in several animal species indicates that
chronic exposure to D-galactose is deleterious to health and the
addition of D-galactose by injections or in the diet is an
established animal model of aging.4-7 Even a low dose of
D-galactose induces changes that resemble natural aging in
animals, including shortened life span caused by oxidative stress
damage, chronic inflammation, neurodegeneration, decreased
immune response, and gene transcriptional changes.5 7 A
subcutaneous dose of 100 mg/kg D-galactose accelerates
senescence in mice.5 This is equivalent to 6-10 g in humans,
corresponding to 1-2 glasses of milk. Based on a concentration
of lactose in cow’s milk of approximately 5%, one glass of milk
comprises about 5 g of D-galactose. The increase of oxidative
stress with aging and chronic low grade inflammation is not

Correspondence to: K Michaëlsson karl.michaelsson@surgsci.uu.se

Extra material supplied by the author (see http://www.bmj.com/content/349/bmj.g6015?tab=related#datasupp)
Supplementary information

No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe: http://www.bmj.com/subscribe

BMJ 2014;349:g6015 doi: 10.1136/bmj.g6015 (Published 27 October 2014)
only a pathogenetic mechanism of cardiovascular disease and
cancer in humans8 9 but also a mechanism of age related bone
loss and sarcopenia.9 10 The high amount of lactose and therefore
D-galactose in milk with theoretical influences on processes
such as oxidative stress and inflammation makes the
recommendations to increase milk intake for prevention of
fractures a conceivable contradiction.
Because of the high content of lactose in milk, we hypothesised
that high consumption of milk may increase oxidative stress,
which in turn affects the risk of mortality and fracture.
Meta-analyses of cohort studies for the association between
dairy and milk intake in relation to mortality11 and fractures12 13
have displayed no clear pattern of risk, and evidence from
randomised trials are lacking. Separating milk intake from the
consumption of other dairy products may be of importance since
a less pronounced induction of oxidative stress and inflammation
in humans is expected with cheese and fermented dairy products
(for example, soured milk and yogurt) because of their lower
or non-existent lactose and galactose content,14 15 possible
probiotic antioxidant and anti-inflammatory effects,16-18 and
effects on gut microbiota.19-21 Indeed, a high intake of fermented
milk products has been associated with a decreased risk of
cardiovascular diseases,18 22-24 whereas a high milk intake is
related to a tendency of an unfavourable risk profile for the
development of diabetes and cardiovascular disease.18 23 24 We
therefore assessed the relation between high milk intake with
risk of death and fractures in women and men. We also studied
biological markers of oxidative stress and inflammation in
relation to milk intake in humans.
Methods
We used two community based cohorts, the Swedish
Mammography Cohort25 26 and the Cohort of Swedish Men,27
to analyse the association of milk consumption and mortality
and fracture rates. Figure 1⇓ shows the study sample. In
1987-90, all 90 303 women aged 39-74 years residing in two
Swedish counties (Uppsala and Västmanland, both in central
Sweden) received a postal invitation to a routine mammography
screening. Enclosed with this invitation was a questionnaire
covering both diet (food frequency questionnaire) and lifestyle,
which was completed by 74% of the women. In 1997, a
subsequent, expanded questionnaire was sent to those who were
still living in the study area (response rate 70%). In the present
study 61 433 women in the Swedish Mammography Cohort
with baseline data from 1987-90 and 38 984 with updated
information from 1997 were available for analysis.
The Cohort of Swedish Men was created in the autumn of 1997.
All men, aged 45-79 years, residing in Örebro and Västmanland
counties in central Sweden were invited to participate in the
study (n=100 303). Enclosed with this invitation was a
questionnaire covering both diet (food frequency questionnaire)
and lifestyle, which was completed by 48 850 men. Despite the
response rate of 49%, the Cohort of Swedish Men is considered
representative of Swedish men in this age range in terms of age
distribution, educational level, and prevalence of being
overweight.27 After exclusions, the final sample included 45
339 men (fig 1).
Food frequency questionnaires
The participants reported their average frequency of
consumption of up to 96 foods and beverages during the past
year,25 27 28—that is, how many servings, a day or a week, they
consumed of common foods, including milk, fermented milk,
yogurt, and cheese. Instructions were given that one serving of
No commercial reuse: See rights and reprints http://www.bmj.com/permissions
Page 2 of 15
RESEARCH
milk corresponded to one glass of 200 mL. In the first
questionnaire in the Swedish Mammography Cohort the
categories were prespecified, but in the second questionnaire
BMJ: first published as 10.1136/bmj.g6015 on 28 October 2014. Downloaded from http://www.bmj.com/ on 1 March 2019 by guest. Protected by copyright.
and the one used in the Cohort of Swedish Men, participants
could fill in the exact number of servings of the dairy products
(milk, fermented milk, yogurt, and cheese) they consumed a
day or a week. Milk intake was specified according to fat
content, and we summed intake into a single measure
representing total milk intake on a continuous scale. We
estimated nutrient intakes by multiplying the consumption
frequency of each food item by the nutrient content of age
specific portion sizes and reference data obtained from the
Swedish National Food Agency database.29 The residual method
was used to adjust all nutrient intakes for total energy intake.30
According to validation studies of milk intake, the correlation
between the food frequency questionnaire and four, seven day
food records every third month, a gold standard reference, has
been approximately 0.7.31 Furthermore, in both sexes we have
found a positive association between reported intake of milk
and the fat tissue content of pentadecanoic acid, a biological
marker reflecting average long term intake of milk fat—that is,
present in both milk and fermented milk products.32 33
Outcomes
We considered outcomes registered between study entry (date
of mammography screening in 1987-90 for the Swedish
Mammography Cohort and 1 January 1998 for the Cohort of
Swedish Men) and 31 December 2010. Follow-up until death
was through the Swedish cause of death registry. Complete
linkage with the register is possible by the personal identity
number provided to all Swedish residents. We used the
underlying cause of death from the Swedish cause of death
registry to define mortality from all causes, cardiovascular
diseases (international classification of diseases, 10th revision;
ICD-10 codes I00–I99), and cancer (ICD-10 C-codes). For 1987
to 1996, we used corresponding ICD-9 codes. Accuracy of
classification of causes of death in the Swedish registry is high.34
We collated fracture events through linkage with the Swedish
national patient registry. We defined any fracture event as a
hospital admission or an outpatient visit with ICD-10 codes
S12-S92. Hip fracture cases were defined by the codes
S720-S722. Using a previously validated and accurate method,
we separated admissions for incident fracture from readmissions
for a previous fracture event.25 35 In analyses we used only the
first fracture event. We retained cases of fractures due to
suspected high impact trauma in the analysis since there are
indications of comparable increases in the risks of low impact
and high impact trauma fractures in association with decreasing
bone density in elderly people (≥60 years).36 37 Pathologic
fractures as a consequence of a tumour were not included as an
outcome.
Comorbidity and other additional information
From the questionnaires we obtained information on lifestyle,
weight, and height. For the Swedish Mammography Cohort the
questionnaires also covered information on use of
postmenopausal oestrogen therapy, menopausal status, and
parity. In the 1997 questionnaires for both cohorts, information
was collected on smoking status, ever use of cortisone, and
leisure time physical activity during the past year, with five
predefined categories ranging from one hour a week to more
than five hours a week. This physical activity assessment is
valid compared with activity records and accelerometer data.38
We divided educational level into four categories: <9 years,
10-12 years, >12 years, and other (such as vocational). Living
Subscribe: http://www.bmj.com/subscribe
BMJ 2014;349:g6015 doi: 10.1136/bmj.g6015 (Published 27 October 2014)
alone was categorised based on marital status (yes: unmarried,
divorced, widow/widower; no: married, cohabiting). To calculate
Charlson’s comorbidity index we collated diagnosis codes from
the national patient registry.39 40
Biomarkers
D-galactose supplementation in animals has been shown to
increase oxidative stress and inflammation.4-7 To assess the
association between milk intake and biological markers of
oxidative stress and inflammation (fig 1), we additionally
analysed a clinical subcohort of the Swedish Mammography
Cohort25 and the previously described Uppsala Longitudinal
Study of Adult Men cohort.41 We assessed food intake by a third
food frequency questionnaire in the Swedish Mammography
Cohort Clinical (n=5022; mean age 70 years) and by recording
diet for one week at age 71 years in the Uppsala Longitudinal
Study of Adult Men (n=1138). In 892 women (mean age 70
years) and 633 men (urine collected at age 77) we analysed the
urine oxidative stress marker 8-iso-PGF2α, a dominant
F2-isoprostane and an ideal standard biomarker of oxidative
stress in vivo.42 We used serum from the same age group in the
Uppsala Longitudinal Study of Adult Men to analyse interleukin
6 (a main inflammatory biomarker, n=700).
Statistical analysis
For each participant we calculated time at risk until date of each
outcome, date of emigration, or the end of the study period (31
December 2010), whichever occurred first. We used Cox
proportional hazards regression for estimation of age adjusted
and multivariable adjusted hazard ratios and their 95%
confidence intervals for prespecified categories of milk intake
(<200, 200-399, 400-599, and ≥600 g/d) and for continuous
milk intake for each 200 g/d, corresponding to one glass of milk.
Because other dairy products may have different health effects
from that of milk,18 23 we additionally studied intake of fermented
milk and cheese as exposures. The proportional hazard
assumptions were confirmed graphically by log-log plots. We
assessed non-linear trends of risk using restricted cubic-spline
Cox regression with three knots placed at centiles 10, 50, and
90 of the milk product intake.43
We used the directed acyclic graph approach44 to select suitable
covariates for the multivariable model (see supplementary
appendix figure A). The model included age, total energy intake,
body mass index, height (all continuous), educational level (≤9,
10-12, >12 years, other), living alone (yes/no), calcium
supplementation (yes/no), vitamin D supplementation (yes/no),
ever use of cortisone (yes/no), a healthy dietary pattern, as
previously described45 (continuous), physical activity (five
categories), smoking status (never, former, current), and the
Charlson comorbidity index (continuous, 1-16). We included
other potential covariates such as menopausal status; hormone
replacement therapy; intakes of calcium, vitamin D, total fat,
saturated fat, retinol, alcohol, potassium, phosphorus, and
protein; parity; and previous fracture of any type (the last when
analysing hip fracture as an outcome, in a second multivariable
model to assess whether these covariates confounded or
mediated potential associations). In the Swedish Mammography
Cohort, of which we had access to information from repeat
questionnaires, we time updated exposures and covariates,
theoretically yielding stronger outcome associations.46 The
analysis in the Cohort of Swedish Men was based on a single
exposure assessment. Using the Markov chain Monte Carlo
multiple imputation methods, we imputed covariates not
assessed in the baseline questionnaire of the Swedish
Mammography Cohort in 1987-90 but assessed in the second
No commercial reuse: See rights and reprints http://www.bmj.com/permissions
Page 3 of 15
RESEARCH
questionnaire in 1997 (for example, smoking status and physical
activity). We performed sensitivity analysis limiting the analysis
to baseline data at 1987-90 and 1997 using ordinary Cox’s
BMJ: first published as 10.1136/bmj.g6015 on 28 October 2014. Downloaded from http://www.bmj.com/ on 1 March 2019 by guest. Protected by copyright.
regression without time updated information. In an additional
sensitivity analysis, we adjusted our estimates for milk intake
by fermented milk and cheese intakes, and vice versa. We also
adjusted our second multivariable model for additional nutrients
(except lactose) known to be constituents of milk. Accordingly,
we therefore evaluated the change of our hazard ratios after
adjustment for intakes of riboflavin, thiamine, niacin, vitamin
B6, vitamin B12, folate, β carotene, iron, magnesium, sodium,
selenium, and zinc.
To assess the influence of competing events, we compared
cumulative incidence curves for hip fracture, treating death from
all causes as competing event, with the Kaplan-Meier curves
for hip fracture by categories of milk intake.47 Even though
competing risk analysis is not considered appropriate for causal
analysis, which was the purpose of our study, it can provide
additional valuable information for risk prediction.48 We
undertook a further sensitivity analysis with any type of vehicle
related non-fracture incident as an outcome (identified by
ICD-10 codes V01-99 excluding simultaneous fracture), which
should theoretically be unrelated to consumption of milk. We
evaluated whether a history of fracture was associated with milk
consumption in the Uppsala Longitudinal Study of Adult Men
cohort and whether experiencing a fracture or comorbidity
between the first and second food frequency questionnaires was
associated with changed milk consumption in the Swedish
Mammography Cohort.
Finally, we used cubic-spline regression analysis to assess the
relation between intake of milk, fermented milk, or yogurt,
cheese, the urine oxidative stress marker 8-iso-PGF2α, and
serum interleukin 6. Before analysis we log transformed both
biomarkers because of their skewed distributions. We adjusted
the estimates for age, body mass index, energy intake, education
(four categories), smoking status (never, former, current), and
physical activity (four categories). The statistical analyses were
performed with STATA 11.2 (StataCorp, College Station, TX)
and SAS, version 9.3 (SAS Institute, Cary, NC).
Results
Table 1⇓ lists the characteristics of the study participants by
categories of milk intake. Mean intake of milk at baseline in
the Swedish Mammography Cohort was 240 g a day and in the
Cohort of Swedish Men was 290 g a day. With increasing
categories of milk intake the reported intake for most other
nutrients, including energy intake, also increased, although
alcohol intake tended to decrease. There were generally small
differences between categories of milk intake in body stature,
nutritional supplement use, marital status, comorbidity,
educational level, smoking status, and physical activity level.
During a median of 22 years of follow-up and 1 231 818 person
years at risk in the Swedish Mammography Cohort, 15 541
women died. The underlying cause of death was cardiovascular
disease in 5278 women and cancer in 3283 women. We
ascertained that 17 252 women had any type of fracture, of
whom 4259 had had a hip fracture. During a median of 13 years
of follow-up and 534 094 person years at risk within the Cohort
of Swedish Men, 10 112 men died (4568 from cardiovascular
causes and 2881 from cancer) and 5379 had any type of fracture;
of these, 1166 men had a hip fracture.
Among women in the Swedish Mammography Cohort, with
analysis based on repeated exposure measurements, we observed
a positive association between milk intake and total mortality
Subscribe: http://www.bmj.com/subscribe
BMJ 2014;349:g6015 doi: 10.1136/bmj.g6015 (Published 27 October 2014)
as well as fracture, especially hip fracture (table 2⇓ and adjusted
spline curves in fig 2⇓). In women, higher rates were observed
for death from all causes (adjusted hazard ratio 1.15, 95%
confidence interval 1.13 to 1.17, for each glass of milk),
cardiovascular disease (1.15, 1.12 to 1.19, for each glass of
milk), and cancer (1.07, 1.02 to 1.11, for each glass of milk)
(table 2 and fig 3⇓). Milk consumption corresponding to three
or more glasses of milk a day (mean 680 g a day) compared
with less than one glass a day (mean 60 g a day), was associated
with a hazard ratio of total mortality of 1.93 (1.80 to 2.06) in
women, with approximately similar estimates for cardiovascular
mortality and somewhat lower for cancer mortality (1.44, 1.23
to 1.69). For women who consumed three or more glasses of
milk a day the hazard ratio for any fracture was 1.16 (1.08 to
1.25) and for hip fracture was 1.60 (1.39 to 1.84).
In an analysis based on a single exposure assessment, men in
the Cohort of Swedish Men also had a higher rate of death with
higher milk consumption (table 2, fig 2). However, the excess
risk was less pronounced than in women, with an adjusted
hazard ratio of 1.10 (95% confidence interval 1.03 to 1.17) for
three or more glasses of milk a day (mean 830 g a day)
compared with less than one glass a day (mean 50 g a day) and
was mainly associated with an increased rate of cardiovascular
death (table 2 and fig 3). No reduction in all fractures or hip
fracture rates with increasing milk intake was observed in men
(fig 2).
Other dairy products
Tables A and B in the supplementary appendix show the baseline
characteristics of the participants by categories of cheese and
fermented milk products (yogurt and other soured milk
products). The distribution of the covariates displayed a similar
pattern to that of milk intake categories. In a sensitivity analysis,
the risk estimates of the outcomes associated with consumption
of cheese or fermented milk products were in the opposite
direction of estimates associated with milk consumption. Thus
women with a high intake of cheese or fermented milk products
compared with women with low intakes had lower mortality
and fracture rates (see supplementary tables C and D). For each
serving the rate of mortality and hip fractures was reduced by
10-15% (P<0.001). Risk reductions in men, based on a single
exposure assessment, were more modest or were non-existent
(see supplementary tables C and D).
Sensitivity analysis
We performed an extended multivariable model to evaluate
whether nutrients known to be associated with osteoporosis or
fracture risk (calcium, vitamin D, phosphorus, fat, protein, and
retinol) influenced our estimates. Adjustment for these nutrients
further strengthened the association between milk intake and
outcomes (see supplementary table E). The associations
remained robust after performing further sensitivity analysis.
Accordingly, our estimates for milk intake were independent
of consumption by fermented milk and cheese (see
supplementary tables F and G). The estimates of fermented milk
and cheese were attenuated when adjusted for milk intake and
each other (see supplementary tables F and G). We further
extended our second multivariable model to evaluate the
influence on our estimates by adjustment for additional nutrients
known to be constituents of milk, but the hazard ratios were
only changed by 4% or less.
To investigate possible bias in the Swedish Mammography
Cohort introduced by using time updated information in the
models we conducted additional sensitivity analyses. Firstly,
No commercial reuse: See rights and reprints http://www.bmj.com/permissions
Page 4 of 15
RESEARCH
we included only baseline data, without updating with
information from the second questionnaire (see supplementary
table H). Secondly, we used the second questionnaire as baseline
BMJ: first published as 10.1136/bmj.g6015 on 28 October 2014. Downloaded from http://www.bmj.com/ on 1 March 2019 by guest. Protected by copyright.
(see supplementary table I). These analyses indicated a higher
risk of death and fracture with high milk intakes, although with
attenuated estimates compared with time updated intakes (table
2).
The reporting of milk consumption did not seem to be influenced
by comorbid conditions (see supplementary table J). Thus
women in the Swedish Mammography Cohort who had
experienced one comorbidity or two comorbidities or more in
the follow-up time between the first and the second food
frequency questionnaire had a change in milk consumption
similar to those who did not experience comorbidities during
the period. Similarly, women who had a fracture before the
second food frequency questionnaire did not increase their milk
consumption. The reporting of fermented milk products was
not affected by comorbidity status (see supplementary table J).
No difference (P=0.31) in milk consumption was noted between
the 13% of Uppsala Longitudinal Study of Adult Men who
reported that at least one parent had had a hip fracture (1.2
glasses of milk/d, 95% confidence interval 1.0 to 1.3) and those
who did not have a family history of hip fracture (1.3 glass of
milk/d, 1.2 to 1.3). We did not find an impact of competing
events from mortality for the association between milk intake
and hip fracture (see supplementary figure B). The incidence
of vehicle related non-fracture incidents as a consequence
(n=1161 women) was not higher among those with a high milk
consumption (hazard ratio 0.96, 95% confidence interval 0.90
to 1.03 for each glass of milk). Finally, the excess rate of
mortality associated with higher milk consumption was also
evident in both sexes without a fracture during follow-up (data
not shown).
Milk intake, oxidative stress, and inflammation
We further investigated whether milk intake was associated
with oxidative stress and inflammation. Milk intake was
positively associated with 8-iso-PGF2α in both sexes, and with
interleukin 6 in men (fig 4⇓). Consumption of fermented milk
products (soured milk and yogurt) indicated a negative relation
with both the oxidative stress and the inflammatory markers
(see supplementary figure C, panel A). No such association was
observed with cheese intake (see supplementary figure C, panel
B).
Discussion
We observed a dose dependent higher rate of both mortality and
fracture in women and a higher rate of mortality in men with
milk intake, a pattern not discerned with other dairy products.
Milk intake was not associated with fracture rate in men. There
were positive associations between milk intake and
concentrations of markers for oxidative stress and inflammation.
Strengths and weaknesses of this study
Our study strengths include the population based prospective
designs in both sexes in a setting with a large range of milk
intake. Using repeat measurements of dietary intake in the
Swedish Mammography Cohort increased the accuracy and
precision of measurements on dietary intake. We traced
outcomes through national healthcare registers and deterministic
record linkage, permitting complete ascertainment. Also, we
adjusted for several important covariates, although residual
confounding cannot be excluded—for example, even though
the results were adjusted for some aspects of socioeconomic
Subscribe: http://www.bmj.com/subscribe
BMJ 2014;349:g6015 doi: 10.1136/bmj.g6015 (Published 27 October 2014)
status (education and marital status), additional nuances may
not have been adequately captured by our observational study
design. The findings were independent of non-lactose nutrients
in milk. However, we did not consider the non-nutritional
contents of dairy products such as persistent organic pollutants49
and heavy metals,50 but these contaminants should be
concentrated not only in milk but also in fermented milk
products, which we found to be associated with reduced rates
of mortality and fracture.
Our results might not apply to people of other ethnic origins,
such as those with a high prevalence of lactose intolerance, or
to children and adolescents. Nutrient concentrations in milk and
other dairy products are variable and depend on factors such as
food fortification, biosynthesis, the animal’s diet, and
physicochemical conditions,51 which might affect the
generalisability of our results. Theoretically, the findings on
fractures might be explained by a reverse causation
phenomenon, where people with a higher predisposition for
osteoporosis may have deliberately increased their milk intake.
We investigated time to first fracture, which reduces the
likelihood of biased estimates. Furthermore, high milk
consumption was also related to higher mortality among those
without a fracture during follow-up. In the analyses we did not
consider fractures caused by metastatic cancer, but cases of
fractures due to suspected high impact trauma were, as
recommended,36 37 retained in the analysis since these fractures
are—as ordinary fragility fractures—also more common in those
with low bone mineral density. The possibility of a reverse
causation theory is also contradicted by the fact that fermented
dairy products were related to a reduced risk of fracture and that
a personal or a family history of hip fracture was not associated
with a higher milk intake. Additionally, the change in average
reported consumption of milk in the Swedish Mammography
Cohort during a long follow-up was not affected by change in
comorbidity status. Furthermore, prospective designs are more
likely to generate non-differential misclassification and thus
attenuate the evaluated association. None the less, we cannot
rule out the possibility that our design or analysis failed to
capture a reverse causation phenomenon.
Comparison with other studies
A high intake of milk is accompanied by a higher energy intake,
as indicated by the baseline characteristics of our participants.
However, results from both cohort studies52 53 and randomised
controlled trials54 55 show that a high intake of dairy products is
not associated with an increase in weight or body mass index
despite a higher intake of energy. These results are in line with
the present study where those who reported a high milk intake
also had higher energy intake but a similar body mass index
compared with women and men with a lower milk intake.
Ecological studies suggest higher mortality rates from fracture
and ischaemic heart disease in countries with high milk
consumption.56 57 Higher milk consumption has also been
suggested to affect the risk of certain cancers and cardiovascular
disease.11 58 59 Meta-analyses of cohort studies on the association
between milk intake and mortality11 and fractures12 13 have shown
significant heterogeneity between studies. In the higher milk
consumption category, for example, some show lower mortality
rates and some show higher. Comparisons between studies are
hampered by differences in exposure range, the pooling of
different milk products as exposure, the method used for dietary
assessment, the general dietary pattern, the prevalence of lactose
intolerance, outcome assessment, and study size. To consider
the totality of evidence concerning the role of milk consumption
in death and fractures, we recommend that these differences
No commercial reuse: See rights and reprints http://www.bmj.com/permissions
Page 5 of 15
RESEARCH
between study designs are considered in future metaregression
analyses of observational studies. Our present investigation
should not be evaluated in isolation and its merits should be
BMJ: first published as 10.1136/bmj.g6015 on 28 October 2014. Downloaded from http://www.bmj.com/ on 1 March 2019 by guest. Protected by copyright.
judged in light of other study findings. The fact that no
randomised trial has examined the effect of milk intake on
incidence of mortality and fractures, long term experimental
evidence is needed to confirm a causal association between
higher milk intake and higher mortality. Another possible
analytical approach might be the use of genetic variation in
lactase persistence using a Mendelian randomisation study
design, but these specific genetic variants are probably weak as
an instrumental variable.60 In addition, the dose-response relation
with milk intake is not readily observed with such a design and
neither is the consideration of type of dairy product consumed.
Previous studies also conclude that gene-environmental
interaction seems to be of importance for this specific gene
expression, and pleiotropic effects cannot be excluded.61 62
Our present study of two cohorts included a larger number of
outcomes than the total number of events included in the
meta-analyses of previous studies.11 13 To counteract the random
misclassification occurring in cohort studies and leading to
conservatively biased risk estimates, studies with a large number
of outcomes are needed and also preferentially repeat dietary
assessments to reduce misclassification of the exposure. Indeed,
we found a stronger association with milk intake and both
mortality and fractures in women than in men. This inequality
might be explained by real sex differences but also by the larger
size of the female cohort study and by the repeat food frequency
questionnaire surveys and time updated analysis. It should be
emphasised that when a single exposure assessment was applied
in the sensitivity analysis of the Swedish Mammography Cohort,
by use of either the first or the second food frequency
questionnaire, the hazard ratios were attenuated compared with
the time dependent exposure analysis and were also similar to
the estimates obtained in the male cohort. Therefore, improved
validity and precision of the exposure is a likely explanation
for the higher risk observed in women.
Comparing milk with other dairy products
Particularly noteworthy is that intake of fermented milk products
such as yogurt and soured milk and cheese were associated with
lower rates of fracture and mortality. Furthermore, we observed
a positive association only between milk intake and markers of
oxidative stress (urine 8-iso-PGF2α) and inflammation (serum
interleukin 6). Previously, we found a negative relation between
bone mineral density and 8-iso-PGF2α.42 63 Interleukin 6 seems
to be causally related to cardiovascular disease64 and may
influence bone loss and osteoporosis.65 Importantly, those who
consume high amounts of non-fermented milk have a more
non-favourable cardiovascular risk factor profile, with higher
blood pressure, lower high density lipoprotein cholesterol levels,
and higher insulin resistance.18 In contrast, intake of cheese and
fermented milk products is related to higher high density
lipoprotein cholesterol levels, less insulin resistance, and a lower
risk of myocardial infarction.18 22-24 In addition, a recent small
randomised cross over study indicated that the intake of a
fermented dairy diet seemed to provide a more favourable
biomarker profile than that of a non-fermented dairy diet.66
Potential mechanism
One potential candidate for the discrepant results for different
types of dairy products is D-galactose content. The intake of
D-galactose from non-fermented milk is considerably higher
than that from other food sources, including cheese and
fermented milk products. Non-dairy sources of D-galactose are
Subscribe: http://www.bmj.com/subscribe
BMJ 2014;349:g6015 doi: 10.1136/bmj.g6015 (Published 27 October 2014)
mainly cereals, vegetables, and fruits,67 but the concentration
of galactose and the amount ingested from these sources
accounts for a small proportion of the total intake of galactose.
Put into perspective, the amount of lactose in one glass of milk
corresponds to approximately 5 g of galactose, whereas the
amount in 100 g of fruits or vegetables67 is measured in
milligrams or tens of milligrams. D-galactose given to laboratory
animals (mice, rats, and drosophila flies) is an established
experimental model for premature aging, including shortened
life span caused by oxidative stress and chronic inflammation,4-7
but whether this mechanism can be generalised to humans needs
further scientific support. However, galactosaemia is a genetic
disorder that results from loss of
galactose-1P-uridylyltransferase, with accumulation of galactose
in blood and other tissues as a consequence.68 69 Affected infants
experience a rapid escalation of potentially lethal acute
symptoms after exposure to milk, and experimental models
display oxidative stress as a mechanism for the development of
disease.68 Even with dietary restrictions of galactose intake these
patients have higher circulating levels of galactose and an
increased risk for chronic diseases in adulthood,69 including
osteoporosis.70
Conclusion
A higher consumption of milk in women and men is not
accompanied by a lower risk of fracture and instead may be
associated with a higher rate of death. Consequently, there may
be a link between the lactose and galactose content of milk and
risk as suggested in our hypothesis, although causality needs
be tested using experimental study designs. Our results may
question the validity of recommendations to consume high
amounts of milk to prevent fragility fractures.3 71 72 The results
should, however, be interpreted cautiously given the
observational design of our study. The findings merit
independent replication before they can be used for dietary
recommendations.
Contributors: KM and LB designed the study and drafted the manuscript.
KM and LB analysed the data. LB, HM, SL, SB, EWL, and AW
contributed to the interpretation of the data and revision of the
manuscript. KM had primary responsibility for final content and acts as
guarantor. All authors read and approved the final manuscript. The
funders of the study had no role in the study design, data collection,
data analysis, data interpretation, writing of the report, or the decision
to submit the article for publication.
Funding: This study was supported by grants from the Swedish Research
Council. EWL is employed by the Swedish National Food Agency. The
views in this article do not necessarily represent those of the agency.
Competing interests: All authors have completed the ICMJE uniform
disclosure form at www.icmje.org/coi_disclosure.pdf and declare: no
support from any organisation for the submitted work; no financial
relationships with any organisations that might have an interest in the
submitted work in the previous three years; no other relationships or
activities that could appear to have influenced the submitted work.
Ethical approval: The Swedish Mammography Cohort, the clinical
subcohort of the Swedish Mammography Cohort, and the Cohort of
Swedish Men were approved by the regional research ethics board at
Karolinska Insitutet. The Uppsala Longitudinal Study of Adult Men was
approved by the regional research ethics board at Uppsala University.
All participants gave informed consent.
Data sharing: Summary data from the Uppsala Longitudinal Study of
Adult Men are available at http://www.pubcare.uu.se/ULSAM.
Transparency: The lead author (KM) affirms that the manuscript is an
honest, accurate, and transparent account of the study being reported;
No commercial reuse: See rights and reprints http://www.bmj.com/permissions
Page 6 of 15
RESEARCH
that no important aspects of the study have been omitted; and that any
discrepancies from the study as planned (and, if relevant, registered)
have been explained.
BMJ: first published as 10.1136/bmj.g6015 on 28 October 2014. Downloaded from http://www.bmj.com/ on 1 March 2019 by guest. Protected by copyright.
1 Burger J, Kirchner M, Bramanti B, Haak W, Thomas MG. Absence of the
lactase-persistence-associated allele in early Neolithic Europeans. Proc Natl Acad Sci U
S A 2007;104:3736-41.
2 Simoons FJ. The geographic hypothesis and lactose malabsorption. A weighing of the
evidence. Am J Dig Dis 1978;23:963-80.
3 Heaney RP. Dairy and bone health. J Am Coll Nutr 2009;28(Suppl 1):82S-90S.
4 Song X, Bao M, Li D, Li YM. Advanced glycation in D-galactose induced mouse aging
5
model. Mech Ageing Dev 1999;108:239-51.
Cui X, Zuo P, Zhang Q, Li X, Hu Y, Long J, et al. Chronic systemic D-galactose exposure
induces memory loss, neurodegeneration, and oxidative damage in mice: protective
effects of R-alpha-lipoic acid. J Neurosci Res 2006;83:1584-90.
6 Hao L, Huang H, Gao J, Marshall C, Chen Y, Xiao M. The influence of gender, age and
treatment time on brain oxidative stress and memory impairment induced by d-galactose
in mice. Neurosci Lett 2014;571C:45-9.
7 Cui X, Wang L, Zuo P, Han Z, Fang Z, Li W, et al. D-galactose-caused life shortening in
Drosophila melanogaster and Musca domestica is associated with oxidative stress.
Biogerontology 2004;5:317-25.
8 Reuter S, Gupta SC, Chaturvedi MM, Aggarwal BB. Oxidative stress, inflammation, and
cancer: how are they linked? Free Radic Biol Med 2010;49:1603-16.
9 Manolagas SC, Parfitt AM. What old means to bone. Trends Endocrinol Metab
10
2010;21:369-74.
Michaëlsson K, Wolk A, Byberg L, Ärnlöv J, Melhus H. Intake and serum concentrations
of alpha-tocopherol in relation to fractures in elderly women and men: 2 cohort studies.
Am J Clin Nutr 2014;99:107-14.
11 Soedamah-Muthu SS, Ding EL, Al-Delaimy WK, Hu FB, Engberink MF, Willett WC, et al.
Milk and dairy consumption and incidence of cardiovascular diseases and all-cause
mortality: dose-response meta-analysis of prospective cohort studies. Am J Clin Nutr
2011;93:158-71.
12 Kanis JA, Johansson H, Oden A, De Laet C, Johnell O, Eisman JA, et al. A meta-analysis
of milk intake and fracture risk: low utility for case finding. Osteoporos Int 2005;16:799-804.
13 Bischoff-Ferrari HA, Dawson-Hughes B, Baron JA, Kanis JA, Orav EJ, Staehelin HB, et
al. Milk intake and risk of hip fracture in men and women: a meta-analysis of prospective
14
cohort studies. J Bone Miner Res 2011;26:833-9.
Alm L. Effect of fermentation on lactose, glucose, and galactose content in milk and
suitability of fermented milk products for lactose intolerant individuals. J Dairy Sci
15
1982;65:346-52.
Portnoi PA, MacDonald A. Determination of the lactose and galactose content of cheese
for use in the galactosaemia diet. J Hum Nutr Diet 2009;22:400-8.
16 Kumar M, Kumar A, Nagpal R, Mohania D, Behare P, Verma V, et al. Cancer-preventing
attributes of probiotics: an update. Int J Food Sci Nutr 2010;61:473-96.
17 Nestel PJ, Mellett N, Pally S, Wong G, Barlow CK, Croft K, et al. Effects of low-fat or
full-fat fermented and non-fermented dairy foods on selected cardiovascular biomarkers
in overweight adults. Br J Nutr 2013:110:2242-9.
18 Sonestedt E, Wirfalt E, Wallstrom P, Gullberg B, Orho-Melander M, Hedblad B. Dairy
products and its association with incidence of cardiovascular disease: the Malmo diet and
19
cancer cohort. Eur J Epidemiol 2011;26:609-18.
Ceapa C, Wopereis H, Rezaiki L, Kleerebezem M, Knol J, Oozeer R. Influence of fermented
milk products, prebiotics and probiotics on microbiota composition and health. Best Pract
Res Clin Gastroenterol 2013;27:139-55.
20 Sommer F, Backhed F. The gut microbiota—masters of host development and physiology.
Nat Rev Microbiol 2013;11:227-38.
21 McNulty NP, Yatsunenko T, Hsiao A, Faith JJ, Muegge BD, Goodman AL, et al. The
impact of a consortium of fermented milk strains on the gut microbiome of gnotobiotic
mice and monozygotic twins. Sci Transl Med 2011;3:106ra06.
22 Astrup A. Yogurt and dairy product consumption to prevent cardiometabolic diseases:
epidemiologic and experimental studies. Am J Clin Nutr 2014;99:1235S-42S.
23 Patterson E, Larsson SC, Wolk A, Akesson A. Association between dairy food consumption
and risk of myocardial infarction in women differs by type of dairy food. J Nutr
2013;143:74-9.
24 Huth PJ, Park KM. Influence of dairy product and milk fat consumption on cardiovascular
disease risk: a review of the evidence. Adv Nutr 2012;3:266-85.
25 Warensjö E, Byberg L, Melhus H, Gedeborg R, Mallmin H, Wolk A, et al. Dietary calcium
intake and risk of fracture and osteoporosis: prospective longitudinal cohort study. BMJ
;342:d1473.
26 Michaëlsson K, Melhus H, Warensjö Lemming E, Wolk A, Byberg L. Long term calcium
intake and rates of all cause and cardiovascular mortality: community based prospective
longitudinal cohort study. BMJ 2013;346:f228.
27 Thomas LD, Michaelsson K, Julin B, Wolk A, Akesson A. Dietary cadmium exposure and
fracture incidence among men: a population-based prospective cohort study. J Bone
Miner Res 2011;26:1601-8.
28 Larsson SC, Bergkvist L, Wolk A. Long-term dietary calcium intake and breast cancer
risk in a prospective cohort of women. Am J Clin Nutr 2009;89:277-82.
29 Bergström L, Kylberg E, Hagman U, Eriksson H, Bruce Å. The food composition database
KOST: the National Food Administration’s information system for nutritive values of food
[Swedish]. Vår föda 1991;43:439-47.
30 Willett WC, Howe GR, Kushi LH. Adjustment for total energy intake in epidemiologic
studies. Am J Clin Nutr 1997;65(4 Suppl):1220S-28S; discussion 29S-31S.
31 Larsson SC, Andersson SO, Johansson JE, Wolk A. Cultured milk, yoghurt, and dairy
intake in relation to bladder cancer risk in a prospective study of Swedish women and
men. Am J Clin Nutr 2008;88:1083-7.
32 Jiang J, Wolk A, Vessby B. Relation between the intake of milk fat and the occurrence of
conjugated linoleic acid in human adipose tissue. Am J Clin Nutr 1999;70:21-7.
33 Wolk A, Vessby B, Ljung H, Barrefors P. Evaluation of a biological marker of dairy fat
intake. Am J Clin Nutr 1998;68:291-5.
34 Tunstall-Pedoe H, Kuulasmaa K, Amouyel P, Arveiler D, Rajakangas AM, Pajak A.
Myocardial infarction and coronary deaths in the World Health Organization MONICA
Project. Registration procedures, event rates, and case-fatality rates in 38 populations
from 21 countries in four continents. Circulation 1994;90:583-612.
Subscribe: http://www.bmj.com/subscribe
BMJ 2014;349:g6015 doi: 10.1136/bmj.g6015 (Published 27 October 2014)
What is already known on this topic
A high milk intake is recommended for the prevention of osteoporotic fractures
Milk is the major dietary source of galactose intake
The addition of galactose by injection or in the diet is an established animal model of aging by induction of oxidative stress and inflammation
Results of previous research on the importance of milk intake for the prevention of fractures and the influence on mortality rates are
conflicting
What this study adds
A high milk intake in both sexes is associated with higher mortality and fracture rates and with higher levels of oxidative stress and
inflammatory biomarkers
Such a pattern was not observed with high intake of fermented milk products
35 Gedeborg R, Engquist H, Berglund L, Michaelsson K. Identification of incident injuries in
hospital discharge registers. Epidemiology 2008;19:860-7.
36 Sanders KM, Pasco JA, Ugoni AM, Nicholson GC, Seeman E, Martin TJ, et al. The
exclusion of high trauma fractures may underestimate the prevalence of bone fragility
fractures in the community: the Geelong Osteoporosis Study. J Bone Miner Res
1998;13:1337-42.
37 Mackey DC, Lui LY, Cawthon PM, Bauer DC, Nevitt MC, Cauley JA, et al. High-trauma
fractures and low bone mineral density in older women and men. JAMA 2007;298:2381-8.
38 Orsini N, Bellocco R, Bottai M, Hagstromer M, Sjostrom M, Pagano M, et al. Validity of
self-reported total physical activity questionnaire among older women. Eur J Epidemiol
2008;23:661-7.
39 Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic
comorbidity in longitudinal studies: development and validation. J Chronic Dis
1987;40:373-83.
40 Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi J, et al. Coding algorithms
for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care
2005;43:1130-9.
41 Michaëlsson K, Lithell H, Vessby B, Melhus H. Serum retinol levels and the risk of fracture.
N Engl J Med 2003;348:287-94.
42 Basu S, Michaëlsson K, Olofsson H, Johansson S, Melhus H. Association between
oxidative stress and bone mineral density. Biochem Biophys Res Commun 2001;288:275-9.
43 STATA Corporation. Stata reference manual, release 11. In: Stata Corporation, ed. Stata
Press, 2009.
44 VanderWeele TJ, Hernan MA, Robins JM. Causal directed acyclic graphs and the direction
of unmeasured confounding bias. Epidemiology 2008;19:720-8.
45 Michaëlsson K, Melhus H, Warensjö Lemming E, Wolk A, Byberg L. Long term calcium
intake and rates of all cause and cardiovascular mortality: community based prospective
longitudinal cohort study. BMJ 2013;346:f228.
46 Hu FB, Stampfer MJ, Rimm E, Ascherio A, Rosner BA, Spiegelman D, et al. Dietary fat
and coronary heart disease: a comparison of approaches for adjusting for total energy
intake and modeling repeated dietary measurements. Am J Epidemiol 1999;149:531-40.
47 Lin DY. Non-parametric inference for cumulative incidence functions in competing risks
studies. Stat Med 1997;16:901-10.
48 Lau B, Cole SR, Gange SJ. Competing risk regression models for epidemiologic data.
Am J Epidemiol 2009;170:244-56.
49 Tornkvist A, Glynn A, Aune M, Darnerud PO, Ankarberg EH. PCDD/F, PCB, PBDE, HBCD
and chlorinated pesticides in a Swedish market basket from 2005—levels and dietary
intake estimations. Chemosphere 2011;83:193-9.
50 Tahvonen R, Kumpulainen J. Lead and cadmium contents in milk, cheese and eggs on
the Finnish market. Food Addit Contam 1995;12:789-98.
51 Gaucheron F. Milk and dairy products: a unique micronutrient combination. J Am Coll
Nutr 2011;30(5 Suppl 1):400S-9S.
52 Louie JC, Flood VM, Hector DJ, Rangan AM, Gill TP. Dairy consumption and overweight
and obesity: a systematic review of prospective cohort studies. Obes Rev 2011;12:e582-92.
53 Mozaffarian D, Hao T, Rimm EB, Willett WC, Hu FB. Changes in diet and lifestyle and
long-term weight gain in women and men. N Engl J Med 2011;364:2392-404.
54 Abargouei AS, Janghorbani M, Salehi-Marzijarani M, Esmaillzadeh A. Effect of dairy
consumption on weight and body composition in adults: a systematic review and
meta-analysis of randomized controlled clinical trials. Int J Obes (Lond) 2012;36:1485-93.
55 Chen M, Pan A, Malik VS, Hu FB. Effects of dairy intake on body weight and fat: a
meta-analysis of randomized controlled trials. Am J Clin Nutr 2012;96:735-47.
No commercial reuse: See rights and reprints http://www.bmj.com/permissions
Page 7 of 15
RESEARCH
BMJ: first published as 10.1136/bmj.g6015 on 28 October 2014. Downloaded from http://www.bmj.com/ on 1 March 2019 by guest. Protected by copyright.
56 Hegsted DM. Fractures, calcium, and the modern diet. Am J Clin Nutr 2001;74:571-3.
57 Segall JJ. Hypothesis: is lactose a dietary risk factor for ischaemic heart disease? Int J
Epidemiol 2008;37:1204-8.
58 Song Y, Chavarro JE, Cao Y, Qiu W, Mucci L, Sesso HD, et al. Whole milk intake is
associated with prostate cancer-specific mortality among U.S. male physicians. J Nutr
2013;143:189-96.
59 Lampe JW. Dairy products and cancer. J Am Coll Nutr 2011;30(5 Suppl 1):464S-70S.
60 Timpson NJ, Brennan P, Gaborieau V, Moore L, Zaridze D, Matveev V, et al. Can lactase
persistence genotype be used to reassess the relationship between renal cell carcinoma
and milk drinking? Potentials and problems in the application of Mendelian randomization.
Cancer Epidemiol Biomarkers Prev 2010;19:1341-8.
61 Corella D, Arregui M, Coltell O, Portoles O, Guillem-Saiz P, Carrasco P, et al. Association
of the LCT-13910C>T polymorphism with obesity and its modulation by dairy products in
a Mediterranean population. Obesity (Silver Spring) 2011;19:1707-14.
62 Wagh K, Bhatia A, Alexe G, Reddy A, Ravikumar V, Seiler M, et al. Lactase persistence
and lipid pathway selection in the Maasai. PLoS One 2012;7:e44751.
63 Östman B, Michaëlsson K, Helmersson J, Byberg L, Gedeborg R, Melhus H, et al.
Oxidative stress and bone mineral density in elderly men: antioxidant activity of
alpha-tocopherol. Free Radic Biol Med 2009;47:668-73.
64 IL6R Genetics Consortium Emerging Risk Factors Collaboration, Sarwar N, Butterworth
AS, Freitag DF, Gregson J, Willeit P, et al. Interleukin-6 receptor pathways in coronary
heart disease: a collaborative meta-analysis of 82 studies. Lancet 2012;379:1205-13.
65 Clowes JA, Riggs BL, Khosla S. The role of the immune system in the pathophysiology
of osteoporosis. Immunol Rev 2005;208:207-27.
66 Nestel PJ, Mellett N, Pally S, Wong G, Barlow CK, Croft K, et al. Effects of low-fat or
full-fat fermented and non-fermented dairy foods on selected cardiovascular biomarkers
in overweight adults. Br J Nutr 2013;110:2242-9.
67 Gross KC, Acosta PB. Fruits and vegetables are a source of galactose: implications in
planning the diets of patients with galactosaemia. J Inherit Metab Dis 1991;14:253-8.
68 Jumbo-Lucioni PP, Hopson ML, Hang D, Liang Y, Jones DP, Fridovich-Keil JL. Oxidative
stress contributes to outcome severity in a Drosophila melanogaster model of classic
galactosemia. Dis Models Mech 2013;6:84-94.
69 Lai K, Elsas LJ, Wierenga KJ. Galactose toxicity in animals. IUBMB life 2009;61:1063-74.
70 Batey LA, Welt CK, Rohr F, Wessel A, Anastasoaie V, Feldman HA, et al. Skeletal health
in adult patients with classic galactosemia. Osteoporos Int 2013;24:501-9.
71 US Department of Agriculture and US Department of Health and Human Services. Dietary
Guidelines for Americans. 2010. www.cnpp.usda.gov/dietaryguidelines.htm.
72 Weaver CM. How sound is the science behind the dietary recommendations for dairy?
Am J Clin Nutr 2014;99:1217S-22S.
Accepted: 22 September 2014
Cite this as: BMJ 2014;349:g6015
This is an Open Access article distributed in accordance with the Creative Commons
Attribution Non Commercial (CC BY-NC 3.0) license, which permits others to distribute,
remix, adapt, build upon this work non-commercially, and license their derivative works
on different terms, provided the original work is properly cited and the use is
non-commercial. See: http://creativecommons.org/licenses/by-nc/3.0/.
Subscribe: http://www.bmj.com/subscribe
BMJ 2014;349:g6015 doi: 10.1136/bmj.g6015 (Published 27 October 2014) Page 8 of 15

RESEARCH

Tables

BMJ: first published as 10.1136/bmj.g6015 on 28 October 2014. Downloaded from http://www.bmj.com/ on 1 March 2019 by guest. Protected by copyright.
Table 1| Baseline characteristics of women in Swedish Mammography Cohort (1987-90) and men in Cohort of Swedish Men (1997) by
categories of milk intake. Values are numbers (percentages) unless stated otherwise

Categories of daily milk intake

Characteristics <1 glass (<200 g/d) 1-2 glasses (200-399 g/d) 2-3 glasses (400-599 g/d) ≥3 glasses (≥600 g/d)
Swedish Mammography Cohort
No 16 926 23 438 15 461 5608
Mean (SD) age at entry (years) 53.2 (9.6) 54.0 (9.7) 54.1 (9.9) 52.8 (9.6)
Mean (SD) body mass index (kg/m2) 24.5 (3.8) 24.7 (3.8) 25.0 (3.9) 24.9 (4.1)
Mean (SD) height (cm) 164.0 (5.7) 164.0 (5.7) 164.0 (5.8) 164.3 (5.9)
Mean (SD) nutrient intake:
Energy (kcal/d) 1412 (433) 1535 (414) 1707 (435) 1965 (527)
Calcium (mg/d) 733 (159) 859 (140) 973 (144) 1101 (175)
Phosphorus (mg/d) 1241 (185) 1365 (165) 1478 (175) 1587 (204)
Vitamin D (mg/d) 3.9 (1.7) 4.4 (1.2) 4.8 (1.4) 5.1 (1.7)
Retinol (mg/d) 0.94 (0.70) 1.03 (0.64) 1.09 (0.61) 1.09 (0.59)
Protein (g/d) 62.2 (9.2) 66.4 (8.0) 69.9 (8.2) 73.1 (8.8)
Total fat (g/d) 44.3 (17.1) 48.9 (16.7) 55.0 (18.5) 64.9 (24.3)
Saturated fat (g/d) 19.0 (8.1) 21.4 (8.0) 24.5 (9.1) 30.0 (12.6)
Alcohol (g/d) 3.1 (3.9) 2.6 (3.5) 2.1 (3.0) 1.9 (3.0)
Mean (SD) metabolic equivalents (kcal/kg and hour) 42.2 (4.8) 42.4 (4.8) 42.6 (4.8) 42.8 (5.0)
Calcium supplement use* 6287 (37.1) 7558 (32.2) 3976 (25.7) 1300 (23.2)
Vitamin D supplement use* 2539 (15.0) 3397 (14.5) 1950 (12.6) 714 (12.7)
Ever oestrogen replacement use* 3502 (20.7) 5066 (21.6) 3518 (22.8) 1464 (26.1)
Education level:
≤9 years 13 284 (78.5) 18 721 (79.9) 12 540 (81.1) 4417 (78.8)
10-12 years 1370 (8.1) 1702 (7.3) 1057 (6.8) 406 (7.2)
>12 years 897 (5.3) 1148 (4.9) 632 (4.1) 266 (4.7)
other 1375 (8.1) 1867 (8.0) 1232 (8.0) 519 (9.2)
Charlson comorbidity:
0 15 197 (89.8) 21 154 (90.3) 13·866 (89.7) 4928 (87.9)
1 1404 (8.3) 1798 (7.7) 1236 (8.0) 533 (9.5)
≥2 325 (1.9) 486 (2.1) 359 (2.3) 147 (2.6)
Nulliparous 2034 (12.0) 2542 (10.8) 1556 (10.1) 566 (10.9)
Ever cortisone use* 914 (5.4) 1096 (4.6) 759 (4.9) 262 (4.7)
Smoker status:*
Current 3353 (19.8) 4295 (18.3) 3096 (20.0) 1346 (24.0)
Former 5340 (31.6) 6890 (29.4) 4459 (28.8) 1656 (29.5)
Never 8233 (48.6) 12 253 (52.3) 7906 (51.1) 2606 (46.5)
Living alone 3958 (23.4) 5308 (22.6) 3761 (24.3) 1441 (25.7)
Cohort of Swedish Men
No 18 459 10 841 8927 7112
Mean (SD) age at entry (years) 58.9 (9.4) 61.0 (9.8) 61.9 (9.9) 60.9 (9.7)
Mean (SD) body mass index (kg/m2) 25.6 (3.2) 25.6 (3.3) 25.8 (3.4) 26.4 (3.6)
Mean (SD) height (cm) 177.4 (6.7) 177.2 (6.6) 177.1 (6.7) 177.2 (6.8)
Mean (SD) nutrient intake:
Energy (kcal/d) 2513 (795) 2567 (769) 2718 (796) 3111 (915)
Calcium (mg/d) 1239 (390) 1420 (363) 1624 (393) 1931 (493)
Phosphorus (mg/d) 1917 (297) 2043 (280) 2174 (295) 2378 (355)

No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe: http://www.bmj.com/subscribe

BMJ 2014;349:g6015 doi: 10.1136/bmj.g6015 (Published 27 October 2014) Page 9 of 15

RESEARCH

Table 1 (continued)

Categories of daily milk intake

BMJ: first published as 10.1136/bmj.g6015 on 28 October 2014. Downloaded from http://www.bmj.com/ on 1 March 2019 by guest. Protected by copyright.
Characteristics <1 glass (<200 g/d) 1-2 glasses (200-399 g/d) 2-3 glasses (400-599 g/d) ≥3 glasses (≥600 g/d)
Vitamin D (mg/d) 6.1 (2.9) 6.6 (2.9) 7.1 (2.8) 7.7 (3.2)
Retinol (mg/d) 1.18 (0.97) 1.24 (0.87) 1.28 (0.72) 1.35 (0.84)
Protein (g/d) 98.3 (14.8) 101.5 (14.0) 104.9 (14.1) 110.9 (15.3)
Total fat (g/d) 89.6 (15.7) 89.2 (14.5) 89.3 (14.9) 89.2 (15.6)
Saturated fat (g/d) 40.2 (9.6) 40.5 (9.0) 41.3 (9.4) 42.2 (10.0)
Alcohol (g/d) 16.1 (22.8) 12.4 (17.5) 11.6 (21.5) 11.3 (20.5)
Metabolic equivalents (kcal/kg and hour) 41.2 (4.8) 41.4 (4.8) 41.8 (5.0) 42.3 (5.3)
Calcium supplement use 1930 (10.5) 1101 (10.2) 833 (9.3) 576 (8.1)
Vitamin D supplement use 2542 (13.8) 1521 (14.0) 1156 (12.9) 834 (11.7)
Educational level:
≤9 years 11 822 (64.3) 7324 (67.8) 6663 (74.9) 5553 (78.4)
10-12 years 2954 (16.1) 1576 (14.6) 1029 (11.6) 714 (10.1)
>12 years 3551 (19.3) 1864 (17.2) 1166 (13.1) 787 (11.1)
other 71 (0.4) 45 (0.4) 33 (0.4) 31 (0.4)
Charlson comorbidity:
0 15 836 (85.8) 8975 (82.8) 7262 (81.3) 5809 (81.7)
1 1865 (10.1) 1304 (12.0) 1148 (12.9) 885 (12.4)
≥2 758 (4.1) 562 (5.2) 517 (5.8) 418 (5.9)
Ever cortisone use 753 (4.1) 456 (4.2) 386 (4.3) 318 (4.5)
Smoker status:
Current 4365 (24.0) 2419 (22.6) 2284 (26.0) 2041 (29.1)
Former 7513 (41.2) 4120 (38.5) 3209 (36.6) 2573 (36.7)
Never 6347 (34.8) 4160 (38.9) 3286 (37.4) 2394 (34.2)
Living alone 3116 (16.9) 1877 (17.3) 1646 (18.4) 1462 (20.6)

*Information available only from 1997 questionnaire; values are imputed from these data.

No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe: http://www.bmj.com/subscribe

BMJ 2014;349:g6015 doi: 10.1136/bmj.g6015 (Published 27 October 2014) Page 10 of 15

RESEARCH

Table 2| Milk consumption and time to death and fracture in Swedish Mammography Cohort* and Cohort of Swedish Men

Categories of daily milk intake

BMJ: first published as 10.1136/bmj.g6015 on 28 October 2014. Downloaded from http://www.bmj.com/ on 1 March 2019 by guest. Protected by copyright.
1-2 glasses (200-399 2-3 glasses (400-599 Continuous (per 200
Variables <1 glass (<200 g/d) g/d) g/d) ≥3 glasses (≥600 g/d) g)
Swedish Mammography Cohort
Total mortality:
No of deaths 5422 5830 3150 1139 15541
Person years of follow-up 476 485 444 724 229648 80961 1 231 818
Rate/1000 person years† 10.7 12.6 15.4 17.7 12.6
Age adjusted HR (95% CI) 1.00 (reference) 1.30 (1.25 to 1.35) 1.83 (1.75 to 1.91) 2.20 (2.06 to 2.35) 1.18 (1.16 to 1.20)
Multivariable adjusted HR (95% CI)‡ 1.00 (reference) 1.21 (1.16 to 1.25) 1.60 (1.53 to 1.68) 1.93 (1.80 to 2.06) 1.15 (1.13 to 1.17)
Cardiovascular mortality:
No of deaths 1904 1989 1030 355 5278
Person years of follow-up 476 485 444 724 229 648 80 961 1 231 818
Rate/1000 person years† 3.6 4.3 5.3 6.2 4.3
Age adjusted HR (95% CI) 1.00 (reference) 1.26 (1.18 to 1.34) 1.85 (1.71 to 1.99) 2.19 (1.96 to 2.46) 1.20 (1.16 to 1.23)
Multivariable adjusted HR (95% CI)‡ 1.00 (reference) 1.16 (1.09 to 1.24) 1.59 (1.47 to 1.73) 1.90 (1.69 to 2.14) 1.15 (1.12 to 1.19)
Cancer mortality:
No of deaths 1412 1196 488 187 3283
Person years of follow-up 476 485 444 724 229 648 80 961 1 231 818
Rate/1000 person years† 2.8 2.6 2.4 2.8 2.7
Age adjusted HR (95% CI) 1.00 (reference) 1.11 (1.02 to 1.19) 1.24 (1.12 to 1.37) 1.55 (1.33 to 1.80) 1.09 (1.05 to 1.13)
Multivariable adjusted HR (95% CI)‡ 1.00 (reference) 1.07 (0.99 to 1.15) 1.16 (1.04 to 1.29) 1.44 (1.23 to 1.69) 1.07 (1.02 to 1.11)
Any fracture:
No of fractures 7080 6317 2912 943 17252
Person years of follow-up 424 857 402 138 210 610 75 231 1 112 837
Rate/1000 person years† 16.0 15.4 15.0 15.1 15.5
Age adjusted HR (95% CI) 1.00 (reference) 1.08 (1.04 to 1.11) 1.17 (1.12 to 1.22) 1.16 (1.08 to 1.24) 1.02 (1.00 to 1.03)
Multivariable adjusted HR (95% CI)‡ 1.00 (reference) 1.07 (1.04 to 1.11) 1.16 (1.11 to 1.21) 1.16 (1.08 to 1.25) 1.02 (1.00 to 1.04)
Hip fracture:
No of fractures 1584 1620 808 247 4259
Person years of follow-up 468 603 436 885 225 969 79 838 1 211 295
Rate/1000 person years† 3.1 3.6 4.1 4.2 3.5
Age adjusted HR (95% CI) 1.00 (reference) 1.24 (1.16 to 1.33) 1.69 (1.55 to 1.85) 1.76 (1.54 to 2.02) 1.11 (1.08 to 1.15)
Multivariable adjusted HR (95% CI)‡ 1.00 (reference) 1.19 (1.11 to 1.28) 1.55 (1.41 to 1.69) 1.60 (1.39 to 1.84) 1.09 (1.05 to 1.13)
Cohort of Swedish Men
Total mortality:
No of deaths 3468 2501 2347 1796 10112
Person years of follow-up 221 381 127 248 103 049 82 415 534094
Rate/1000 person years† 18.2 18.3 19.6 20.7 18.9
Age adjusted HR (95% CI) 1.00 (reference) 1.00 (0.95 to 1.05) 1.07 (1.02 to 1.13) 1.13 (1.07 to 1.19) 1.03 (1.02 to 1.05)
Multivariable adjusted HR (95% CI)‡ 1.00 (reference) 0.99 (0.94 to 1.05) 1.05 (1.00 to 1.11) 1.10 (1.03 to 1.17) 1.03 (1.01 to 1.04)
Cardiovascular mortality:
No of deaths 1468 1161 1098 841 4568
Person years of follow-up 221 381 127 248 103049 82 415 534 094
Rate/1000 person years† 7.9 8.4 9.0 9.6 8.6
Age adjusted HR (95% CI) 1.00 (reference) 1.06 (0.98 to 1.15) 1.14 (1.05 to 1.23) 1.21 (1.11 to 1.32) 1.05 (1.03 to 1.07)
Multivariable adjusted HR (95% CI)‡ 1.00 (reference) 1.04 (0.96 to 1.12) 1.10 (1.01 to 1.19) 1.16 (1.06 to 1.27) 1.05 (1.03 to 1.07)
Cancer mortality:
No of deaths 1077 704 616 484 2881
Person years of follow-up 221 381 127 248 103 049 82 415 534 094

No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe: http://www.bmj.com/subscribe

BMJ 2014;349:g6015 doi: 10.1136/bmj.g6015 (Published 27 October 2014) Page 11 of 15

RESEARCH

Table 2 (continued)

Categories of daily milk intake

BMJ: first published as 10.1136/bmj.g6015 on 28 October 2014. Downloaded from http://www.bmj.com/ on 1 March 2019 by guest. Protected by copyright.
1-2 glasses (200-399 2-3 glasses (400-599 Continuous (per 200
Variables <1 glass (<200 g/d) g/d) g/d) ≥3 glasses (≥600 g/d) g)
Rate/1000 person years† 5.5 5.2 5.4 5.6 5.4
Age adjusted HR (95% CI) 1.00 (reference) 0.95 (0.87 to 1.05) 0.97 (0.88 to 1.07) 1.03 (0.92 to 1.14) 1.00 (0.98 to 1.03)
Multivariable adjusted HR (95% CI)‡ 1.00 (reference) 0.97 (0.88 to 1.07) 0.97 (0.87 to 1.07) 1.01 (0.90 to 1.13) 0.99 (0.97 to 1.02)
Any fracture:
No of fractures 2095 1326 1095 863 5379
Person years of follow-up 211 554 121 162 98 290 78 355 509 361
Rate/1000 person years† 10.5 10.7 10.5 10.8 10.6
Age adjusted HR (95% CI) 1.00 (reference) 1.02 (0.95 to 1.10) 1.01 (0.94 to 1.09) 1.03 (0.96 to 1.12) 1.01 (1.00 to 1.03)
Multivariable adjusted HR (95% CI)‡ 1.00 (reference) 1.02 (0.96 to 1.10) 1.01 (0.93 to 1.08) 1.03 (0.94 to 1.11) 1.01 (0.99 to 1.03)
Hip fracture:
No of fractures 439 309 318 200 1266
Person years of follow-up 219 925 126 241 102 126 81 755 530 047
Rate/1000 person years† 2.4 2.3 2.7 2.3 2.4
Age adjusted HR (95% CI) 1.00 (reference) 0.95 (0.82 to 1.10) 1.12 (0.96 to 1.29) 0.97 (0.82 to 1.15) 1.02 (0.98 to 1.06)
Multivariable adjusted HR (95% CI)‡ 1.00 (reference) 0.95 (0.82 to 1.11) 1.13 (0.97 to 1.31) 1.01 (0.85 to 1.20) 1.03 (0.99 to 1.07)

HR=hazard ratio.
*Had access to repeat exposure information; exposures and covariates were treated as cumulative averages.
†Age standardised rates, number of cases per 1000 person years at risk.
‡Adjusted for age, body mass index, height, total energy intake, total alcohol intake, healthy dietary pattern, calcium and vitamin D supplementation, ever use of
cortisone, educational level, living alone, physical activity level estimated as metabolic equivalents, smoking status, and Charlson’s comorbidity index; and in
women additionally for use of oestrogen replacement therapy and nulliparity.

No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe: http://www.bmj.com/subscribe

BMJ 2014;349:g6015 doi: 10.1136/bmj.g6015 (Published 27 October 2014) Page 12 of 15

RESEARCH

Figures

BMJ: first published as 10.1136/bmj.g6015 on 28 October 2014. Downloaded from http://www.bmj.com/ on 1 March 2019 by guest. Protected by copyright.
Fig 1 Flow chart of study samples

No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe: http://www.bmj.com/subscribe

BMJ 2014;349:g6015 doi: 10.1136/bmj.g6015 (Published 27 October 2014) Page 13 of 15

RESEARCH

BMJ: first published as 10.1136/bmj.g6015 on 28 October 2014. Downloaded from http://www.bmj.com/ on 1 March 2019 by guest. Protected by copyright.
Fig 2 Sex specific multivariable adjusted spline curves of relation between milk intake with time to death from all causes,
hip fracture, and any type of fracture. Covariates were age, total energy intake, body mass index, height, educational level,
living alone, calcium supplementation, vitamin D supplementation, ever use of cortisone, healthy dietary pattern, physical
activity, smoking status, and Charlson’s comorbidity index. The spike plot represents the distribution of milk intake. One
glass of milk corresponds to 200 g

No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe: http://www.bmj.com/subscribe

BMJ 2014;349:g6015 doi: 10.1136/bmj.g6015 (Published 27 October 2014) Page 14 of 15

RESEARCH

BMJ: first published as 10.1136/bmj.g6015 on 28 October 2014. Downloaded from http://www.bmj.com/ on 1 March 2019 by guest. Protected by copyright.
Fig 3 Sex specific multivariable adjusted spline curves of relation between milk intake with time to death from all cancer
and cardiovascular disease (CVD). Covariates were age, total energy intake, body mass index, height, educational level,
living alone, calcium supplementation, vitamin D supplementation, ever use of cortisone, healthy dietary pattern, physical
activity, smoking status, and Charlson’s comorbidity index. The spike plot represents the distribution of milk intake. One
glass of milk corresponds to 200 g

No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe: http://www.bmj.com/subscribe

BMJ 2014;349:g6015 doi: 10.1136/bmj.g6015 (Published 27 October 2014) Page 15 of 15

RESEARCH

BMJ: first published as 10.1136/bmj.g6015 on 28 October 2014. Downloaded from http://www.bmj.com/ on 1 March 2019 by guest. Protected by copyright.
Fig 4 Adjusted predictions of urine log(8-iso-PGF2α), a marker of oxidative stress, in 892 women (based on cross sectional
data, mean age 70 years) and 700 men, and serum log(interleukin 6), a marker of inflammation, in 633 men after cubic-spline
regression with milk consumption. Data for men are based on milk consumption assessed at age 71 years and measurement
of inflammatory markers at age 77 years. Covariates were age, body mass index, energy intake, education, smoking status,
and physical activity. One glass of milk corresponds to 200 g

No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe: http://www.bmj.com/subscribe